Dataset Information

Addressing census data problems in race imputation via fully Bayesian Improved Surname Geocoding and name supplements.

ABSTRACT: Prediction of individuals' race and ethnicity plays an important role in studies of racial disparity. Bayesian Improved Surname Geocoding (BISG), which relies on detailed census information, has emerged as a leading methodology for this prediction task. Unfortunately, BISG suffers from two data problems. First, the census often contains zero counts for minority groups in the locations where members of those groups reside. Second, many surnames-especially those of minorities-are missing from the census data. We introduce a fully Bayesian BISG (fBISG) methodology that accounts for census measurement error by extending the naïve Bayesian inference of the BISG methodology. We also use additional data on last, first, and middle names taken from the voter files of six Southern states where self-reported race is available. Our empirical validation shows that the fBISG methodology and name supplements substantially improve the accuracy of race imputation, especially for racial minorities.

SUBMITTER: Imai K

PROVIDER: S-EPMC11318662 | biostudies-literature | 2022 Dec

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Addressing census data problems in race imputation via fully Bayesian Improved Surname Geocoding and name supplements.

Imai Kosuke K Olivella Santiago S Rosenman Evan T R ETR

Science advances 20221209 49

Prediction of individuals' race and ethnicity plays an important role in studies of racial disparity. Bayesian Improved Surname Geocoding (BISG), which relies on detailed census information, has emerged as a leading methodology for this prediction task. Unfortunately, BISG suffers from two data problems. First, the census often contains zero counts for minority groups in the locations where members of those groups reside. Second, many surnames-especially those of minorities-are missing from the ...[more]

PMID: 36490334

Similar Datasets

Project description:ObjectiveRacial and ethnic disparities and/or inequities have been documented in traffic safety research. However, race/ethnicity data are often not captured in population-level traffic safety databases, limiting the field's ability to comprehensively study racial/ethnic differences in transportation outcomes, as well as our ability to mitigate them. To overcome this limitation, we explored the utility of estimating race and ethnicity for drivers in the New Jersey Safety and Health Outcomes (NJ-SHO) data warehouse using the Bayesian Improved Surname Geocoding (BISG) algorithm. In addition, we summarize important recommendations established to guide researchers developing and implementing racial and ethnic disparity research.MethodsWe applied BISG to estimate population-level race/ethnicity for New Jersey drivers in 2017 and evaluated the concordance between reported values available in integrated administrative sources (e.g., hospital records) and BISG probability distributions using an area under the receiver operator curve (AUC) within each race/ethnicity category. Overall AUC was calculated by weighting each AUC value by the population count in each reported category. In an exemplar analysis using 2017 crash data, we conducted an analysis of average monthly police-reported crash rates in 2017 by race/ethnicity using the NJ-SHO and BISG sets of race/ethnicity values to compare their outputs.ResultsWe found excellent or outstanding concordance (AUC ≥0.86) between reported race/ethnicity and BISG probabilities for White, Hispanic, Black, and Asian/Pacific Islander drivers. We found poor concordance for American Indian/Alaskan Native drivers (AUC= 0.65), and concordance was no better than random assignment for Multiracial drivers (AUC = 0.52). Among White, Hispanic, Asian/Pacific Islander, and American Indian/Alaskan native drivers, monthly crash rates calculated using both NJ-SHO reported race/ethnicity values and BISG probabilities were similar. Monthly crash rates differed by 11% for Black drivers, and by more than 200% for Multiracial drivers.ConclusionFindings of excellent or outstanding concordance between and mostly similar crash rates derived from reported race/ethnicity and BISG probabilities for White, Hispanic, Black, and Asian/Pacific Islander drivers (98.9% of all drivers in this sample) demonstrate the potential utility of BISG in enabling research on transportation disparities and inequities. Concordance between race/ethnicity values were not acceptable for American Indian/Alaskan Native and Multiracial drivers, which is similar to previous applications and evaluations of BISG. Future work is needed to determine the extent to which BISG may be applied to traffic safety contexts.

Project description:BackgroundQuestionnaires used in longitudinal studies may have questions added or removed over time for numerous reasons. Data missing completely at a follow-up survey is a unique issue for longitudinal studies. While such excluded questions lack information at one follow-up survey, they are collected at other follow-up surveys, and covariances observed at other follow-up surveys may allow for the recovery of the missing data. This study utilized data from a large longitudinal cohort study to assess the efficiency and feasibility of using multiple imputation (MI) to recover this type of information.MethodsMillennium Cohort Study participants completed the 9-item Patient Health Questionnaire (PHQ) depression module at 2 time points (2004, 2007). The suicidal ideation item in the module was set to missing for the 2007 assessment. Several single-level MI models using different sets of predictors and forms of suicidal ideation were used to compare self-reported values and imputed values for this item in 2007. Additionally, associations with sleep duration and smoking status, which are related constructs, were compared between self-reported and imputed values of suicidal ideation.ResultsAmong 63,028 participants eligible for imputation analysis, 4.05% reported suicidal ideation on the 2007 survey. The imputation models successfully identified suicidal ideation, with a sensitivity ranging between 34 and 66% and a positive predictive value between 36 and 42%. Specificity remained above 96% and negative predictive value above 97% for all imputed models. Similar associations were found for all imputation models on related constructs, though the dichotomous suicidal ideation imputed from the model using only PHQ depression items yielded estimates that were closest with the self-reported associations for all adjusted analyses.ConclusionsAlthough sensitivity and positive predictive value were relatively low, applying MI techniques allowed for inclusion of an otherwise missing variable. Additionally, correlations with related constructs were estimated near self-reported values. Therefore, the other 8 depression items can be used to estimate suicidal ideation that was completely missing from a survey using MI. However, these imputed values should not be used to estimate population prevalence.

Project description:BackgroundRace is an important predictor of TKA outcomes in the United States; however, analyses of race can be confounded by socioeconomic factors, which can result in difficulty determining the root cause of disparate outcomes after TKA.Questions/purposesWe asked: (1) Are race and socioeconomic factors at the individual level associated with patient-reported pain and function 2 years after TKA? (2) What is the interaction between race and community poverty and patient-reported pain and function 2 years after TKA?MethodsWe identified all patients undergoing TKA enrolled in a hospital-based registry between 2007 and 2011 who provided 2-year outcomes and lived in New York, Connecticut, or New Jersey. Of patients approached to participate in the registry, more than 82% consented and provided baseline data, and of these patients, 72% provided 2-year data. Proportions of patients with complete followup at 2 years were lower among blacks (57%) than whites (74%), among patients with Medicaid insurance (51%) compared with patients without Medicaid insurance (72%), and among patients without a college education (67%) compared with those with a college education (71%). Our final study cohort consisted of 4035 patients, 3841 (95%) of whom were white and 194 (5%) of whom were black. Using geocoding, we linked individual-level registry data to US census tracts data through patient addresses. We constructed a multivariate linear mixed-effect model in multilevel frameworks to assess the interaction between race and census tract poverty on WOMAC pain and function scores 2 years after TKA. We defined a clinically important effect as 10 points on the WOMAC (which is scaled from 1 to 100 points, with higher scores being better).ResultsRace, education, patient expectations, and baseline WOMAC scores are all associated with 2-year WOMAC pain and function; however, the effect sizes were small, and below the threshold of clinical importance. Whites and blacks from census tracts with less than 10% poverty have similar levels of pain and function 2 years after TKA (WOMAC pain, 1.01 ± 1.59 points lower for blacks than for whites, p = 0.53; WOMAC function, 2.32 ± 1.56 lower for blacks than for whites, p = 0.14). WOMAC pain and function scores 2 years after TKA worsen with increasing levels of community poverty, but do so to a greater extent among blacks than whites. Disparities in pain and function between blacks and whites are evident only in the poorest communities; decreasing in a linear fashion as poverty increases. In census tracts with greater than 40% poverty, blacks score 6 ± 3 points lower (worse) than whites for WOMAC pain (p = 0.03) and 7 ± 3 points lower than whites for WOMAC function (p = 0.01).ConclusionsBlacks and whites living in communities with little poverty have similar patient-reported TKA outcomes, whereas in communities with high levels of poverty, there are important racial disparities. Efforts to improve TKA outcomes among blacks will need to address individual- and community-level socioeconomic factors.Level of evidenceLevel III, therapeutic study.

Dataset Information

Addressing census data problems in race imputation via fully Bayesian Improved Surname Geocoding and name supplements.

Publications

Addressing census data problems in race imputation via fully Bayesian Improved Surname Geocoding and name supplements.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets